Spaces:
Running
Running
Upload 3 files
Browse files- dataloader/dataloader.py +593 -0
- dataloader/meldataset.py +383 -0
- dataloader/misc.py +111 -0
dataloader/dataloader.py
ADDED
|
@@ -0,0 +1,593 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import numpy as np
|
| 2 |
+
import math, os, csv
|
| 3 |
+
import torchaudio
|
| 4 |
+
import torch
|
| 5 |
+
import torch.nn as nn
|
| 6 |
+
import torch.utils.data as data
|
| 7 |
+
import torch.distributed as dist
|
| 8 |
+
import soundfile as sf
|
| 9 |
+
from torch.utils.data import Dataset
|
| 10 |
+
import torch.utils.data as data
|
| 11 |
+
import os
|
| 12 |
+
import sys
|
| 13 |
+
sys.path.append(os.path.dirname(__file__))
|
| 14 |
+
from pydub import AudioSegment
|
| 15 |
+
from dataloader.misc import read_and_config_file, get_file_extension
|
| 16 |
+
import librosa
|
| 17 |
+
import random
|
| 18 |
+
EPS = 1e-6
|
| 19 |
+
MAX_WAV_VALUE_16B = 32768.0
|
| 20 |
+
MAX_WAV_VALUE_32B = 2147483648.0
|
| 21 |
+
|
| 22 |
+
def audioread_archieved(path, sampling_rate):
|
| 23 |
+
"""
|
| 24 |
+
Reads an audio file from the specified path, normalizes the audio,
|
| 25 |
+
resamples it to the desired sampling rate (if necessary), and ensures it is single-channel.
|
| 26 |
+
|
| 27 |
+
Parameters:
|
| 28 |
+
path (str): The file path of the audio file to be read.
|
| 29 |
+
sampling_rate (int): The target sampling rate for the audio.
|
| 30 |
+
|
| 31 |
+
Returns:
|
| 32 |
+
numpy.ndarray: The processed audio data, normalized, resampled (if necessary),
|
| 33 |
+
and converted to mono (if the input audio has multiple channels).
|
| 34 |
+
"""
|
| 35 |
+
|
| 36 |
+
# Read audio data and its sample rate from the file.
|
| 37 |
+
data, fs = sf.read(path)
|
| 38 |
+
|
| 39 |
+
# convert to mono channel
|
| 40 |
+
if len(data.shape) >1:
|
| 41 |
+
if data.shape[0] > data.shape[1]:
|
| 42 |
+
data = data[:, 0]
|
| 43 |
+
else:
|
| 44 |
+
data = data[0, :]
|
| 45 |
+
|
| 46 |
+
# Normalize the audio data.
|
| 47 |
+
data, scalar = audio_norm(data)
|
| 48 |
+
|
| 49 |
+
# Resample the audio if the sample rate is different from the target sampling rate.
|
| 50 |
+
if fs != sampling_rate:
|
| 51 |
+
data = librosa.resample(data, orig_sr=fs, target_sr=sampling_rate)
|
| 52 |
+
|
| 53 |
+
# Convert to mono by selecting the first channel if the audio has multiple channels.
|
| 54 |
+
if len(data.shape) > 1:
|
| 55 |
+
data = data[:, 0]
|
| 56 |
+
|
| 57 |
+
# Return the processed audio data.
|
| 58 |
+
return data, scalar
|
| 59 |
+
|
| 60 |
+
def read_audio(file_path):
|
| 61 |
+
"""
|
| 62 |
+
Use AudioSegment to load audio from all supported audio input format
|
| 63 |
+
"""
|
| 64 |
+
|
| 65 |
+
try:
|
| 66 |
+
audio = AudioSegment.from_file(file_path)
|
| 67 |
+
return audio
|
| 68 |
+
except Exception as e:
|
| 69 |
+
print(f"Error loading file: {e}")
|
| 70 |
+
return None
|
| 71 |
+
|
| 72 |
+
def audioread(path, sampling_rate, use_norm):
|
| 73 |
+
"""
|
| 74 |
+
Reads an audio file from the specified path, normalizes the audio,
|
| 75 |
+
resamples it to the desired sampling rate (if necessary), and ensures it is single-channel.
|
| 76 |
+
|
| 77 |
+
Parameters:
|
| 78 |
+
path (str): The file path of the audio file to be read.
|
| 79 |
+
sampling_rate (int): The target sampling rate for the audio.
|
| 80 |
+
use_norm (bool): The flag for specifying whether using input audio normalization
|
| 81 |
+
|
| 82 |
+
Returns:
|
| 83 |
+
numpy.ndarray: The processed audio data, normalized, resampled (if necessary),
|
| 84 |
+
and converted to mono (if the input audio has multiple channels).
|
| 85 |
+
"""
|
| 86 |
+
|
| 87 |
+
# Read audio data and its sample rate from the file.
|
| 88 |
+
audio_info = {}
|
| 89 |
+
ext = get_file_extension(path).replace('.', '')
|
| 90 |
+
audio_info['ext']=ext
|
| 91 |
+
|
| 92 |
+
try:
|
| 93 |
+
data = AudioSegment.from_file(path)
|
| 94 |
+
except Exception as e:
|
| 95 |
+
print(f"Error loading file: {e}")
|
| 96 |
+
return None
|
| 97 |
+
|
| 98 |
+
data = read_audio(path)
|
| 99 |
+
|
| 100 |
+
audio_info['sample_rate'] = data.frame_rate
|
| 101 |
+
audio_info['channels'] = data.channels
|
| 102 |
+
audio_info['sample_width'] = data.sample_width
|
| 103 |
+
|
| 104 |
+
data_array = np.array(data.get_array_of_samples())
|
| 105 |
+
if max(data_array) > MAX_WAV_VALUE_16B:
|
| 106 |
+
audio_np = data_array / MAX_WAV_VALUE_32B
|
| 107 |
+
else:
|
| 108 |
+
audio_np = data_array / MAX_WAV_VALUE_16B
|
| 109 |
+
|
| 110 |
+
audios = []
|
| 111 |
+
# Check if the audio is stereo
|
| 112 |
+
if audio_info['channels'] == 2:
|
| 113 |
+
audios.append(audio_np[::2]) # Even indices (left channel)
|
| 114 |
+
audios.append(audio_np[1::2]) # Odd indices (right channel)
|
| 115 |
+
else:
|
| 116 |
+
audios.append(audio_np)
|
| 117 |
+
|
| 118 |
+
# Normalize the audio data.
|
| 119 |
+
audios_normed = []
|
| 120 |
+
scalars = []
|
| 121 |
+
for audio in audios:
|
| 122 |
+
if use_norm:
|
| 123 |
+
audio_normed, scalar = audio_norm(audio)
|
| 124 |
+
audios_normed.append(audio_normed)
|
| 125 |
+
scalars.append(scalar)
|
| 126 |
+
else:
|
| 127 |
+
audios_normed.append(audio)
|
| 128 |
+
scalars.append(1)
|
| 129 |
+
# Resample the audio if the sample rate is different from the target sampling rate.
|
| 130 |
+
if audio_info['sample_rate'] != sampling_rate:
|
| 131 |
+
index = 0
|
| 132 |
+
for audio_normed in audios_normed:
|
| 133 |
+
audios_normed[index] = librosa.resample(audio_normed, orig_sr=audio_info['sample_rate'], target_sr=sampling_rate)
|
| 134 |
+
index = index + 1
|
| 135 |
+
|
| 136 |
+
# Return the processed audio data.
|
| 137 |
+
return audios_normed, scalars, audio_info
|
| 138 |
+
|
| 139 |
+
def audio_norm(x):
|
| 140 |
+
"""
|
| 141 |
+
Normalizes the input audio signal to a target Root Mean Square (RMS) level,
|
| 142 |
+
applying two stages of scaling. This ensures the audio signal is neither too quiet
|
| 143 |
+
nor too loud, keeping its amplitude consistent.
|
| 144 |
+
|
| 145 |
+
Parameters:
|
| 146 |
+
x (numpy.ndarray): Input audio signal to be normalized.
|
| 147 |
+
|
| 148 |
+
Returns:
|
| 149 |
+
numpy.ndarray: Normalized audio signal.
|
| 150 |
+
"""
|
| 151 |
+
|
| 152 |
+
# Compute the root mean square (RMS) of the input audio signal.
|
| 153 |
+
rms = (x ** 2).mean() ** 0.5
|
| 154 |
+
|
| 155 |
+
# Calculate the scalar to adjust the signal to the target level (-25 dB).
|
| 156 |
+
scalar = 10 ** (-25 / 20) / (rms + EPS)
|
| 157 |
+
|
| 158 |
+
# Scale the input audio by the computed scalar.
|
| 159 |
+
x = x * scalar
|
| 160 |
+
|
| 161 |
+
# Compute the power of the scaled audio signal.
|
| 162 |
+
pow_x = x ** 2
|
| 163 |
+
|
| 164 |
+
# Calculate the average power of the audio signal.
|
| 165 |
+
avg_pow_x = pow_x.mean()
|
| 166 |
+
|
| 167 |
+
# Compute RMS only for audio segments with higher-than-average power.
|
| 168 |
+
rmsx = pow_x[pow_x > avg_pow_x].mean() ** 0.5
|
| 169 |
+
|
| 170 |
+
# Calculate another scalar to further normalize based on higher-power segments.
|
| 171 |
+
scalarx = 10 ** (-25 / 20) / (rmsx + EPS)
|
| 172 |
+
|
| 173 |
+
# Apply the second scalar to the audio.
|
| 174 |
+
x = x * scalarx
|
| 175 |
+
|
| 176 |
+
# Return the doubly normalized audio signal.
|
| 177 |
+
return x, 1/(scalar * scalarx + EPS)
|
| 178 |
+
|
| 179 |
+
class DataReader(object):
|
| 180 |
+
"""
|
| 181 |
+
A class for reading audio data from a list of files, normalizing it,
|
| 182 |
+
and extracting features for further processing. It supports extracting
|
| 183 |
+
features from each file, reshaping the data, and returning metadata
|
| 184 |
+
like utterance ID and data length.
|
| 185 |
+
|
| 186 |
+
Parameters:
|
| 187 |
+
args: Arguments containing the input path and target sampling rate.
|
| 188 |
+
|
| 189 |
+
Attributes:
|
| 190 |
+
file_list (list): A list of audio file paths to process.
|
| 191 |
+
sampling_rate (int): The target sampling rate for audio files.
|
| 192 |
+
"""
|
| 193 |
+
|
| 194 |
+
def __init__(self, args):
|
| 195 |
+
# Read and configure the file list from the input path provided in the arguments.
|
| 196 |
+
# The file list is decoded, if necessary.
|
| 197 |
+
self.file_list = read_and_config_file(args, args.input_path, decode=True)
|
| 198 |
+
|
| 199 |
+
# Store the target sampling rate.
|
| 200 |
+
self.sampling_rate = args.sampling_rate
|
| 201 |
+
|
| 202 |
+
# Store the args file
|
| 203 |
+
self.args = args
|
| 204 |
+
|
| 205 |
+
def __len__(self):
|
| 206 |
+
"""
|
| 207 |
+
Returns the number of audio files in the file list.
|
| 208 |
+
|
| 209 |
+
Returns:
|
| 210 |
+
int: Number of files to process.
|
| 211 |
+
"""
|
| 212 |
+
return len(self.file_list)
|
| 213 |
+
|
| 214 |
+
def __getitem__(self, index):
|
| 215 |
+
"""
|
| 216 |
+
Retrieves the features of the audio file at the given index.
|
| 217 |
+
|
| 218 |
+
Parameters:
|
| 219 |
+
index (int): Index of the file in the file list.
|
| 220 |
+
|
| 221 |
+
Returns:
|
| 222 |
+
tuple: Features (inputs, utterance ID, data length) for the selected audio file.
|
| 223 |
+
"""
|
| 224 |
+
if self.args.task == 'target_speaker_extraction':
|
| 225 |
+
if self.args.network_reference.cue== 'lip':
|
| 226 |
+
return self.file_list[index]
|
| 227 |
+
return self.extract_feature(self.file_list[index])
|
| 228 |
+
|
| 229 |
+
def extract_feature(self, path):
|
| 230 |
+
"""
|
| 231 |
+
Extracts features from the given audio file path.
|
| 232 |
+
|
| 233 |
+
Parameters:
|
| 234 |
+
path (str): The file path of the audio file.
|
| 235 |
+
|
| 236 |
+
Returns:
|
| 237 |
+
inputs (numpy.ndarray): Reshaped audio data for further processing.
|
| 238 |
+
utt_id (str): The unique identifier of the audio file, usually the filename.
|
| 239 |
+
length (int): The length of the original audio data.
|
| 240 |
+
"""
|
| 241 |
+
# Extract the utterance ID from the file path (usually the filename).
|
| 242 |
+
utt_id = path.split('/')[-1]
|
| 243 |
+
use_norm = False
|
| 244 |
+
|
| 245 |
+
#We suggest to use norm for 'FRCRN_SE_16K' and 'MossFormer2_SS_16K' models
|
| 246 |
+
if self.args.network in ['FRCRN_SE_16K','MossFormer2_SS_16K'] :
|
| 247 |
+
use_norm = True
|
| 248 |
+
|
| 249 |
+
# Read and normalize the audio data, converting it to float32 for processing.
|
| 250 |
+
audios_norm, scalars, audio_info = audioread(path, self.sampling_rate, use_norm)
|
| 251 |
+
|
| 252 |
+
if self.args.network in ['MossFormer2_SR_48K']:
|
| 253 |
+
audio_info['sample_rate'] = self.sampling_rate
|
| 254 |
+
|
| 255 |
+
for i in range(len(audios_norm)):
|
| 256 |
+
audios_norm[i] = audios_norm[i].astype(np.float32)
|
| 257 |
+
# Reshape the data to ensure it's in the format [1, data_length].
|
| 258 |
+
audios_norm[i] = np.reshape(audios_norm[i], [1, audios_norm[i].shape[0]])
|
| 259 |
+
|
| 260 |
+
# Return the reshaped audio data, utterance ID, and the length of the original data.
|
| 261 |
+
return audios_norm, utt_id, audios_norm[0].shape[1], scalars, audio_info
|
| 262 |
+
|
| 263 |
+
class Wave_Processor(object):
|
| 264 |
+
"""
|
| 265 |
+
A class for processing audio data, specifically for reading input and label audio files,
|
| 266 |
+
segmenting them into fixed-length segments, and applying padding or trimming as necessary.
|
| 267 |
+
|
| 268 |
+
Methods:
|
| 269 |
+
process(path, segment_length, sampling_rate):
|
| 270 |
+
Processes audio data by reading, padding, or segmenting it to match the specified segment length.
|
| 271 |
+
|
| 272 |
+
Parameters:
|
| 273 |
+
path (dict): A dictionary containing file paths for 'inputs' and 'labels' audio files.
|
| 274 |
+
segment_length (int): The desired length of audio segments to extract.
|
| 275 |
+
sampling_rate (int): The target sampling rate for reading the audio files.
|
| 276 |
+
"""
|
| 277 |
+
|
| 278 |
+
def process(self, path, segment_length, sampling_rate):
|
| 279 |
+
"""
|
| 280 |
+
Reads input and label audio files, and ensures the audio is segmented into
|
| 281 |
+
the desired length, padding if necessary or extracting random segments if
|
| 282 |
+
the audio is longer than the target segment length.
|
| 283 |
+
|
| 284 |
+
Parameters:
|
| 285 |
+
path (dict): Dictionary containing the paths to 'inputs' and 'labels' audio files.
|
| 286 |
+
segment_length (int): Desired length of the audio segment in samples.
|
| 287 |
+
sampling_rate (int): Target sample rate for the audio.
|
| 288 |
+
|
| 289 |
+
Returns:
|
| 290 |
+
tuple: A pair of numpy arrays representing the processed input and label audio,
|
| 291 |
+
either padded to the segment length or trimmed.
|
| 292 |
+
"""
|
| 293 |
+
# Read the input and label audio files using the target sampling rate.
|
| 294 |
+
wave_inputs = audioread(path['inputs'], sampling_rate)
|
| 295 |
+
wave_labels = audioread(path['labels'], sampling_rate)
|
| 296 |
+
|
| 297 |
+
# Get the length of the label audio (assumed both inputs and labels have similar lengths).
|
| 298 |
+
len_wav = wave_labels.shape[0]
|
| 299 |
+
|
| 300 |
+
# If the input audio is shorter than the desired segment length, pad it with zeros.
|
| 301 |
+
if wave_inputs.shape[0] < segment_length:
|
| 302 |
+
# Create zero-padded arrays for inputs and labels.
|
| 303 |
+
padded_inputs = np.zeros(segment_length, dtype=np.float32)
|
| 304 |
+
padded_labels = np.zeros(segment_length, dtype=np.float32)
|
| 305 |
+
|
| 306 |
+
# Copy the original audio into the padded arrays.
|
| 307 |
+
padded_inputs[:wave_inputs.shape[0]] = wave_inputs
|
| 308 |
+
padded_labels[:wave_labels.shape[0]] = wave_labels
|
| 309 |
+
else:
|
| 310 |
+
# Randomly select a start index for segmenting the audio if it's longer than the segment length.
|
| 311 |
+
st_idx = random.randint(0, len_wav - segment_length)
|
| 312 |
+
|
| 313 |
+
# Extract a segment of the desired length from the inputs and labels.
|
| 314 |
+
padded_inputs = wave_inputs[st_idx:st_idx + segment_length]
|
| 315 |
+
padded_labels = wave_labels[st_idx:st_idx + segment_length]
|
| 316 |
+
|
| 317 |
+
# Return the processed (padded or segmented) input and label audio.
|
| 318 |
+
return padded_inputs, padded_labels
|
| 319 |
+
|
| 320 |
+
class Fbank_Processor(object):
|
| 321 |
+
"""
|
| 322 |
+
A class for processing input audio data into mel-filterbank (Fbank) features,
|
| 323 |
+
including the computation of delta and delta-delta features.
|
| 324 |
+
|
| 325 |
+
Methods:
|
| 326 |
+
process(inputs, args):
|
| 327 |
+
Processes the raw audio input and returns the mel-filterbank features
|
| 328 |
+
along with delta and delta-delta features.
|
| 329 |
+
"""
|
| 330 |
+
|
| 331 |
+
def process(self, inputs, args):
|
| 332 |
+
# Convert frame length and shift from seconds to milliseconds.
|
| 333 |
+
frame_length = int(args.win_len / args.sampling_rate * 1000)
|
| 334 |
+
frame_shift = int(args.win_inc / args.sampling_rate * 1000)
|
| 335 |
+
|
| 336 |
+
# Set up configuration for the mel-filterbank computation.
|
| 337 |
+
fbank_config = {
|
| 338 |
+
"dither": 1.0,
|
| 339 |
+
"frame_length": frame_length,
|
| 340 |
+
"frame_shift": frame_shift,
|
| 341 |
+
"num_mel_bins": args.num_mels,
|
| 342 |
+
"sample_frequency": args.sampling_rate,
|
| 343 |
+
"window_type": args.win_type
|
| 344 |
+
}
|
| 345 |
+
|
| 346 |
+
# Convert the input audio to a FloatTensor and scale it to match the expected input range.
|
| 347 |
+
inputs = torch.FloatTensor(inputs * MAX_WAV_VALUE)
|
| 348 |
+
|
| 349 |
+
# Compute the mel-filterbank features using Kaldi's fbank function.
|
| 350 |
+
fbank = torchaudio.compliance.kaldi.fbank(inputs.unsqueeze(0), **fbank_config)
|
| 351 |
+
|
| 352 |
+
# Add delta and delta-delta features.
|
| 353 |
+
fbank_tr = torch.transpose(fbank, 0, 1)
|
| 354 |
+
fbank_delta = torchaudio.functional.compute_deltas(fbank_tr)
|
| 355 |
+
fbank_delta_delta = torchaudio.functional.compute_deltas(fbank_delta)
|
| 356 |
+
fbank_delta = torch.transpose(fbank_delta, 0, 1)
|
| 357 |
+
fbank_delta_delta = torch.transpose(fbank_delta_delta, 0, 1)
|
| 358 |
+
|
| 359 |
+
# Concatenate the original Fbank, delta, and delta-delta features.
|
| 360 |
+
fbanks = torch.cat([fbank, fbank_delta, fbank_delta_delta], dim=1)
|
| 361 |
+
|
| 362 |
+
return fbanks.numpy()
|
| 363 |
+
|
| 364 |
+
class AudioDataset(Dataset):
|
| 365 |
+
"""
|
| 366 |
+
A dataset class for loading and processing audio data from different data types
|
| 367 |
+
(train, validation, test). Supports audio processing and feature extraction
|
| 368 |
+
(e.g., waveform processing, Fbank feature extraction).
|
| 369 |
+
|
| 370 |
+
Parameters:
|
| 371 |
+
args: Arguments containing dataset configuration (paths, sampling rate, etc.).
|
| 372 |
+
data_type (str): The type of data to load (train, val, test).
|
| 373 |
+
"""
|
| 374 |
+
|
| 375 |
+
def __init__(self, args, data_type):
|
| 376 |
+
self.args = args
|
| 377 |
+
self.sampling_rate = args.sampling_rate
|
| 378 |
+
|
| 379 |
+
# Read the list of audio files based on the data type.
|
| 380 |
+
if data_type == 'train':
|
| 381 |
+
self.wav_list = read_and_config_file(args.tr_list)
|
| 382 |
+
elif data_type == 'val':
|
| 383 |
+
self.wav_list = read_and_config_file(args.cv_list)
|
| 384 |
+
elif data_type == 'test':
|
| 385 |
+
self.wav_list = read_and_config_file(args.tt_list)
|
| 386 |
+
else:
|
| 387 |
+
print(f'Data type: {data_type} is unknown!')
|
| 388 |
+
|
| 389 |
+
# Initialize processors for waveform and Fbank features.
|
| 390 |
+
self.wav_processor = Wave_Processor()
|
| 391 |
+
self.fbank_processor = Fbank_Processor()
|
| 392 |
+
|
| 393 |
+
# Clip data to a fixed segment length based on the sampling rate and max length.
|
| 394 |
+
self.segment_length = self.sampling_rate * self.args.max_length
|
| 395 |
+
print(f'No. {data_type} files: {len(self.wav_list)}')
|
| 396 |
+
|
| 397 |
+
def __len__(self):
|
| 398 |
+
# Return the number of audio files in the dataset.
|
| 399 |
+
return len(self.wav_list)
|
| 400 |
+
|
| 401 |
+
def __getitem__(self, index):
|
| 402 |
+
# Get the input and label paths from the list.
|
| 403 |
+
data_info = self.wav_list[index]
|
| 404 |
+
|
| 405 |
+
# Process the waveform inputs and labels.
|
| 406 |
+
inputs, labels = self.wav_processor.process(
|
| 407 |
+
{'inputs': data_info['inputs'], 'labels': data_info['labels']},
|
| 408 |
+
self.segment_length,
|
| 409 |
+
self.sampling_rate
|
| 410 |
+
)
|
| 411 |
+
|
| 412 |
+
# Optionally load Fbank features if specified.
|
| 413 |
+
if self.args.load_fbank is not None:
|
| 414 |
+
fbanks = self.fbank_processor.process(inputs, self.args)
|
| 415 |
+
return inputs * MAX_WAV_VALUE, labels * MAX_WAV_VALUE, fbanks
|
| 416 |
+
|
| 417 |
+
return inputs, labels
|
| 418 |
+
|
| 419 |
+
def zero_pad_concat(self, inputs):
|
| 420 |
+
"""
|
| 421 |
+
Concatenates a list of input arrays, applying zero-padding as needed to ensure
|
| 422 |
+
they all match the length of the longest input.
|
| 423 |
+
|
| 424 |
+
Parameters:
|
| 425 |
+
inputs (list of numpy arrays): List of input arrays to be concatenated.
|
| 426 |
+
|
| 427 |
+
Returns:
|
| 428 |
+
numpy.ndarray: A zero-padded array with concatenated inputs.
|
| 429 |
+
"""
|
| 430 |
+
|
| 431 |
+
# Get the maximum length among all inputs.
|
| 432 |
+
max_t = max(inp.shape[0] for inp in inputs)
|
| 433 |
+
|
| 434 |
+
# Determine the shape of the output based on the input dimensions.
|
| 435 |
+
shape = None
|
| 436 |
+
if len(inputs[0].shape) == 1:
|
| 437 |
+
shape = (len(inputs), max_t)
|
| 438 |
+
elif len(inputs[0].shape) == 2:
|
| 439 |
+
shape = (len(inputs), max_t, inputs[0].shape[1])
|
| 440 |
+
|
| 441 |
+
# Initialize an array with zeros to hold the concatenated inputs.
|
| 442 |
+
input_mat = np.zeros(shape, dtype=np.float32)
|
| 443 |
+
|
| 444 |
+
# Copy the input data into the zero-padded array.
|
| 445 |
+
for e, inp in enumerate(inputs):
|
| 446 |
+
if len(inp.shape) == 1:
|
| 447 |
+
input_mat[e, :inp.shape[0]] = inp
|
| 448 |
+
elif len(inp.shape) == 2:
|
| 449 |
+
input_mat[e, :inp.shape[0], :] = inp
|
| 450 |
+
|
| 451 |
+
return input_mat
|
| 452 |
+
|
| 453 |
+
def collate_fn_2x_wavs(data):
|
| 454 |
+
"""
|
| 455 |
+
A custom collate function for combining batches of waveform input and label pairs.
|
| 456 |
+
|
| 457 |
+
Parameters:
|
| 458 |
+
data (list): List of tuples (inputs, labels).
|
| 459 |
+
|
| 460 |
+
Returns:
|
| 461 |
+
tuple: Batched inputs and labels as torch.FloatTensors.
|
| 462 |
+
"""
|
| 463 |
+
inputs, labels = zip(*data)
|
| 464 |
+
x = torch.FloatTensor(inputs)
|
| 465 |
+
y = torch.FloatTensor(labels)
|
| 466 |
+
return x, y
|
| 467 |
+
|
| 468 |
+
def collate_fn_2x_wavs_fbank(data):
|
| 469 |
+
"""
|
| 470 |
+
A custom collate function for combining batches of waveform inputs, labels, and Fbank features.
|
| 471 |
+
|
| 472 |
+
Parameters:
|
| 473 |
+
data (list): List of tuples (inputs, labels, fbanks).
|
| 474 |
+
|
| 475 |
+
Returns:
|
| 476 |
+
tuple: Batched inputs, labels, and Fbank features as torch.FloatTensors.
|
| 477 |
+
"""
|
| 478 |
+
inputs, labels, fbanks = zip(*data)
|
| 479 |
+
x = torch.FloatTensor(inputs)
|
| 480 |
+
y = torch.FloatTensor(labels)
|
| 481 |
+
z = torch.FloatTensor(fbanks)
|
| 482 |
+
return x, y, z
|
| 483 |
+
|
| 484 |
+
class DistributedSampler(data.Sampler):
|
| 485 |
+
"""
|
| 486 |
+
Sampler for distributed training. Divides the dataset among multiple replicas (processes),
|
| 487 |
+
ensuring that each process gets a unique subset of the data. It also supports shuffling
|
| 488 |
+
and managing epochs.
|
| 489 |
+
|
| 490 |
+
Parameters:
|
| 491 |
+
dataset (Dataset): The dataset to sample from.
|
| 492 |
+
num_replicas (int): Number of processes participating in the training.
|
| 493 |
+
rank (int): Rank of the current process.
|
| 494 |
+
shuffle (bool): Whether to shuffle the data or not.
|
| 495 |
+
seed (int): Random seed for reproducibility.
|
| 496 |
+
"""
|
| 497 |
+
|
| 498 |
+
def __init__(self, dataset, num_replicas=None, rank=None, shuffle=True, seed=0):
|
| 499 |
+
if num_replicas is None:
|
| 500 |
+
if not dist.is_available():
|
| 501 |
+
raise RuntimeError("Requires distributed package to be available")
|
| 502 |
+
num_replicas = dist.get_world_size()
|
| 503 |
+
if rank is None:
|
| 504 |
+
if not dist.is_available():
|
| 505 |
+
raise RuntimeError("Requires distributed package to be available")
|
| 506 |
+
rank = dist.get_rank()
|
| 507 |
+
|
| 508 |
+
self.dataset = dataset
|
| 509 |
+
self.num_replicas = num_replicas
|
| 510 |
+
self.rank = rank
|
| 511 |
+
self.epoch = 0
|
| 512 |
+
self.num_samples = int(math.ceil(len(self.dataset) * 1.0 / self.num_replicas))
|
| 513 |
+
self.total_size = self.num_samples * self.num_replicas
|
| 514 |
+
self.shuffle = shuffle
|
| 515 |
+
self.seed = seed
|
| 516 |
+
|
| 517 |
+
def __iter__(self):
|
| 518 |
+
# Shuffle the indices based on the epoch and seed.
|
| 519 |
+
if self.shuffle:
|
| 520 |
+
g = torch.Generator()
|
| 521 |
+
g.manual_seed(self.seed + self.epoch)
|
| 522 |
+
ind = torch.randperm(int(len(self.dataset) / self.num_replicas), generator=g) * self.num_replicas
|
| 523 |
+
indices = []
|
| 524 |
+
for i in range(self.num_replicas):
|
| 525 |
+
indices = indices + (ind + i).tolist()
|
| 526 |
+
else:
|
| 527 |
+
indices = list(range(len(self.dataset)))
|
| 528 |
+
|
| 529 |
+
# Add extra samples to make the dataset evenly divisible.
|
| 530 |
+
indices += indices[:(self.total_size - len(indices))]
|
| 531 |
+
assert len(indices) == self.total_size
|
| 532 |
+
|
| 533 |
+
# Subsample for the current process.
|
| 534 |
+
indices = indices[self.rank * self.num_samples:(self.rank + 1) * self.num_samples]
|
| 535 |
+
assert len(indices) == self.num_samples
|
| 536 |
+
|
| 537 |
+
return iter(indices)
|
| 538 |
+
|
| 539 |
+
def __len__(self):
|
| 540 |
+
return self.num_samples
|
| 541 |
+
|
| 542 |
+
def set_epoch(self, epoch):
|
| 543 |
+
self.epoch = epoch
|
| 544 |
+
|
| 545 |
+
def get_dataloader(args, data_type):
|
| 546 |
+
"""
|
| 547 |
+
Creates and returns a data loader and sampler for the specified dataset type (train, validation, or test).
|
| 548 |
+
|
| 549 |
+
Parameters:
|
| 550 |
+
args (Namespace): Configuration arguments containing details such as batch size, sampling rate,
|
| 551 |
+
network type, and whether distributed training is used.
|
| 552 |
+
data_type (str): The type of dataset to load ('train', 'val', 'test').
|
| 553 |
+
|
| 554 |
+
Returns:
|
| 555 |
+
sampler (DistributedSampler or None): The sampler for distributed training, or None if not used.
|
| 556 |
+
generator (DataLoader): The PyTorch DataLoader for the specified dataset.
|
| 557 |
+
"""
|
| 558 |
+
|
| 559 |
+
# Initialize the dataset based on the given arguments and dataset type (train, val, or test).
|
| 560 |
+
datasets = AudioDataset(args=args, data_type=data_type)
|
| 561 |
+
|
| 562 |
+
# Create a distributed sampler if distributed training is enabled; otherwise, use no sampler.
|
| 563 |
+
sampler = DistributedSampler(
|
| 564 |
+
datasets,
|
| 565 |
+
num_replicas=args.world_size, # Number of replicas in distributed training.
|
| 566 |
+
rank=args.local_rank # Rank of the current process.
|
| 567 |
+
) if args.distributed else None
|
| 568 |
+
|
| 569 |
+
# Select the appropriate collate function based on the network type.
|
| 570 |
+
if args.network == 'FRCRN_SE_16K' or args.network == 'MossFormerGAN_SE_16K':
|
| 571 |
+
# Use the collate function for two-channel waveform data (inputs and labels).
|
| 572 |
+
collate_fn = collate_fn_2x_wavs
|
| 573 |
+
elif args.network == 'MossFormer2_SE_48K':
|
| 574 |
+
# Use the collate function for waveforms along with Fbank features.
|
| 575 |
+
collate_fn = collate_fn_2x_wavs_fbank
|
| 576 |
+
else:
|
| 577 |
+
# Print an error message if the network type is unknown.
|
| 578 |
+
print(f'in dataloader, please specify a correct network type using args.network!')
|
| 579 |
+
return
|
| 580 |
+
|
| 581 |
+
# Create a DataLoader with the specified dataset, batch size, and worker configuration.
|
| 582 |
+
generator = data.DataLoader(
|
| 583 |
+
datasets,
|
| 584 |
+
batch_size=args.batch_size, # Batch size for training.
|
| 585 |
+
shuffle=(sampler is None), # Shuffle the data only if no sampler is used.
|
| 586 |
+
collate_fn=collate_fn, # Use the selected collate function for batching data.
|
| 587 |
+
num_workers=args.num_workers, # Number of workers for data loading.
|
| 588 |
+
sampler=sampler # Use the distributed sampler if applicable.
|
| 589 |
+
)
|
| 590 |
+
|
| 591 |
+
# Return both the sampler and DataLoader (generator).
|
| 592 |
+
return sampler, generator
|
| 593 |
+
|
dataloader/meldataset.py
ADDED
|
@@ -0,0 +1,383 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import math
|
| 2 |
+
import os
|
| 3 |
+
import random
|
| 4 |
+
import torch
|
| 5 |
+
import torch.utils.data
|
| 6 |
+
import numpy as np
|
| 7 |
+
from librosa.util import normalize
|
| 8 |
+
from scipy.io.wavfile import read
|
| 9 |
+
import scipy
|
| 10 |
+
import librosa
|
| 11 |
+
import wave
|
| 12 |
+
from pydub import AudioSegment
|
| 13 |
+
|
| 14 |
+
MAX_WAV_VALUE = 32768.0
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
def load_wav(full_path):
|
| 18 |
+
try:
|
| 19 |
+
sampling_rate, data = read(full_path)
|
| 20 |
+
if max(data.shape) / sampling_rate < 0.5:
|
| 21 |
+
return None, None
|
| 22 |
+
except FileNotFoundError:
|
| 23 |
+
print(f"File not found: {file_path}")
|
| 24 |
+
return None, None
|
| 25 |
+
except Exception as e:
|
| 26 |
+
print(f"An unexpected error occurred: {e}")
|
| 27 |
+
return None, None
|
| 28 |
+
|
| 29 |
+
if len(data.shape) > 1:
|
| 30 |
+
if data.shape[1] <= 2:
|
| 31 |
+
data = data[...,0]
|
| 32 |
+
else:
|
| 33 |
+
data = data[0,...]
|
| 34 |
+
return data / MAX_WAV_VALUE, sampling_rate
|
| 35 |
+
|
| 36 |
+
def get_wave_duration(file_path):
|
| 37 |
+
"""
|
| 38 |
+
Gets the duration of a WAV file in seconds.
|
| 39 |
+
|
| 40 |
+
:param file_path: Path to the WAV file.
|
| 41 |
+
:return: Duration of the WAV file in seconds.
|
| 42 |
+
"""
|
| 43 |
+
try:
|
| 44 |
+
with wave.open(file_path, 'rb') as wf:
|
| 45 |
+
# Get the number of frames
|
| 46 |
+
num_frames = wf.getnframes()
|
| 47 |
+
# Get the frame rate
|
| 48 |
+
frame_rate = wf.getframerate()
|
| 49 |
+
# Calculate duration
|
| 50 |
+
duration = num_frames / float(frame_rate)
|
| 51 |
+
return duration, frame_rate, num_frames
|
| 52 |
+
except wave.Error as e:
|
| 53 |
+
print(f"Error reading {file_path}: {e}")
|
| 54 |
+
return None, None, None
|
| 55 |
+
except FileNotFoundError:
|
| 56 |
+
print(f"File not found: {file_path}")
|
| 57 |
+
return None, None, None
|
| 58 |
+
except Exception as e:
|
| 59 |
+
print(f"An unexpected error occurred: {e}")
|
| 60 |
+
return None, None, None
|
| 61 |
+
|
| 62 |
+
def read_audio_segment(file_path, start_ms, end_ms):
|
| 63 |
+
"""
|
| 64 |
+
Reads a segment from a WAV file and returns the raw data and its properties.
|
| 65 |
+
|
| 66 |
+
:param file_path: Path to the WAV file.
|
| 67 |
+
:param start_ms: Start time of the segment in milliseconds.
|
| 68 |
+
:param end_ms: End time of the segment in milliseconds.
|
| 69 |
+
:return: A tuple containing the raw audio data, frame rate, sample width, and number of channels.
|
| 70 |
+
"""
|
| 71 |
+
#start_time = time.time()
|
| 72 |
+
try:
|
| 73 |
+
# Load the audio file
|
| 74 |
+
audio = AudioSegment.from_wav(file_path)
|
| 75 |
+
# Extract the segment
|
| 76 |
+
segment = audio[start_ms:end_ms]
|
| 77 |
+
# Get raw audio data
|
| 78 |
+
raw_data = segment.raw_data
|
| 79 |
+
# Get audio properties
|
| 80 |
+
frame_rate = segment.frame_rate
|
| 81 |
+
sample_width = segment.sample_width
|
| 82 |
+
channels = segment.channels
|
| 83 |
+
# Create NumPy array from the raw audio data
|
| 84 |
+
audio_array = np.frombuffer(raw_data, dtype=np.int16)
|
| 85 |
+
|
| 86 |
+
# If stereo, reshape the array to have a second dimension
|
| 87 |
+
if channels > 1:
|
| 88 |
+
audio_array = audio_array.reshape((-1, channels))
|
| 89 |
+
audio_array = audio_array[...,0]
|
| 90 |
+
'''
|
| 91 |
+
if frame_rate !=48000:
|
| 92 |
+
audio_array = audio_array/MAX_WAV_VALUE
|
| 93 |
+
audio_array = librosa.resample(audio_array, frame_rate, 48000)
|
| 94 |
+
audio_array = audio_array * MAX_WAV_VALUE
|
| 95 |
+
frame_rate = 48000
|
| 96 |
+
'''
|
| 97 |
+
#end_time = time.time()
|
| 98 |
+
#time_taken = end_time - start_time
|
| 99 |
+
|
| 100 |
+
#print(f"Successfully read segment from {start_ms}ms to {end_ms}ms in {time_taken:.4f} seconds")
|
| 101 |
+
return audio_array / MAX_WAV_VALUE#, frame_rate #, sample_width, channels
|
| 102 |
+
except Exception as e:
|
| 103 |
+
print(f"An error occurred: {e}")
|
| 104 |
+
return None#, None #, None, None
|
| 105 |
+
|
| 106 |
+
def resample(audio, sr_in, sr_out, target_len=None):
|
| 107 |
+
#audio = audio / MAX_WAV_VALUE
|
| 108 |
+
#audio = normalize(audio) * 0.95
|
| 109 |
+
if target_len is not None:
|
| 110 |
+
audio = scipy.signal.resample(audio, target_len)
|
| 111 |
+
return audio
|
| 112 |
+
resample_factor = sr_out / sr_in
|
| 113 |
+
new_samples = int(len(audio) * resample_factor)
|
| 114 |
+
audio = scipy.signal.resample(audio, new_samples)
|
| 115 |
+
return audio
|
| 116 |
+
|
| 117 |
+
def load_segment(full_path, target_sampling_rate=None, segment_size=None):
|
| 118 |
+
|
| 119 |
+
if segment_size is not None:
|
| 120 |
+
dur,sampling_rate,len_data = get_wave_duration(full_path)
|
| 121 |
+
if sampling_rate is None: return None, None
|
| 122 |
+
if sampling_rate < 44100: return None, None
|
| 123 |
+
|
| 124 |
+
target_dur = segment_size / target_sampling_rate
|
| 125 |
+
if dur < target_dur:
|
| 126 |
+
data, sampling_rate = load_wav(full_path)
|
| 127 |
+
#print(f'data_read: {data.shape}, sampling_rate: {sampling_rate}')
|
| 128 |
+
if data is None: return None, None
|
| 129 |
+
|
| 130 |
+
if target_sampling_rate is not None and sampling_rate != target_sampling_rate:
|
| 131 |
+
data = resample(data, sampling_rate, target_sampling_rate)
|
| 132 |
+
sampling_rate = target_sampling_rate
|
| 133 |
+
data = torch.FloatTensor(data)
|
| 134 |
+
data = data.unsqueeze(0)
|
| 135 |
+
data = torch.nn.functional.pad(data, (0, segment_size - data.size(1)), 'constant')
|
| 136 |
+
data = data.squeeze(0)
|
| 137 |
+
return data.numpy(), sampling_rate
|
| 138 |
+
else:
|
| 139 |
+
dur,sampling_rate,len_data = get_wave_duration(full_path)
|
| 140 |
+
if sampling_rate < 44100: return None, None
|
| 141 |
+
|
| 142 |
+
target_dur = segment_size / target_sampling_rate
|
| 143 |
+
target_len = int(target_dur * sampling_rate)
|
| 144 |
+
start_idx = random.randint(0, (len_data - target_len))
|
| 145 |
+
start_ms = start_idx / sampling_rate * 1000
|
| 146 |
+
end_ms = start_ms + target_dur * 1000
|
| 147 |
+
data = read_audio_segment(full_path, start_ms, end_ms)
|
| 148 |
+
#print(f'data_read: {data.shape}, sampling_rate: {sampling_rate}')
|
| 149 |
+
if data is None: return None, None
|
| 150 |
+
if target_sampling_rate is not None and sampling_rate != target_sampling_rate:
|
| 151 |
+
data = resample(data, sampling_rate, target_sampling_rate)
|
| 152 |
+
sampling_rate = target_sampling_rate
|
| 153 |
+
if len(data) < segment_size:
|
| 154 |
+
data = torch.FloatTensor(data)
|
| 155 |
+
data = data.unsqueeze(0)
|
| 156 |
+
data = torch.nn.functional.pad(data, (0, segment_size - data.size(1)), 'constant')
|
| 157 |
+
data = data.squeeze(0)
|
| 158 |
+
data = data.numpy()
|
| 159 |
+
else:
|
| 160 |
+
start_idx = random.randint(0, (len(data) - segment_size))
|
| 161 |
+
data = data[start_idx:start_idx+segment_size]
|
| 162 |
+
#print(f'data_cut: {data.shape}')
|
| 163 |
+
return data, sampling_rate
|
| 164 |
+
else:
|
| 165 |
+
dur,sampling_rate,len_data = get_wave_duration(full_path)
|
| 166 |
+
if sampling_rate is None: return None, None
|
| 167 |
+
if sampling_rate < 44100: return None, None
|
| 168 |
+
data, sampling_rate = load_wav(full_path)
|
| 169 |
+
if data is None: return None, None
|
| 170 |
+
if target_sampling_rate is not None and sampling_rate != target_sampling_rate:
|
| 171 |
+
data = resample(data, sampling_rate, target_sampling_rate)
|
| 172 |
+
sampling_rate = target_sampling_rate
|
| 173 |
+
return data, sampling_rate
|
| 174 |
+
|
| 175 |
+
def dynamic_range_compression(x, C=1, clip_val=1e-5):
|
| 176 |
+
return np.log(np.clip(x, a_min=clip_val, a_max=None) * C)
|
| 177 |
+
|
| 178 |
+
|
| 179 |
+
def dynamic_range_decompression(x, C=1):
|
| 180 |
+
return np.exp(x) / C
|
| 181 |
+
|
| 182 |
+
|
| 183 |
+
def dynamic_range_compression_torch(x, C=1, clip_val=1e-5):
|
| 184 |
+
return torch.log(torch.clamp(x, min=clip_val) * C)
|
| 185 |
+
|
| 186 |
+
|
| 187 |
+
def dynamic_range_decompression_torch(x, C=1):
|
| 188 |
+
return torch.exp(x) / C
|
| 189 |
+
|
| 190 |
+
|
| 191 |
+
def spectral_normalize_torch(magnitudes):
|
| 192 |
+
output = dynamic_range_compression_torch(magnitudes)
|
| 193 |
+
return output
|
| 194 |
+
|
| 195 |
+
|
| 196 |
+
def spectral_de_normalize_torch(magnitudes):
|
| 197 |
+
output = dynamic_range_decompression_torch(magnitudes)
|
| 198 |
+
return output
|
| 199 |
+
|
| 200 |
+
|
| 201 |
+
mel_basis = {}
|
| 202 |
+
hann_window = {}
|
| 203 |
+
|
| 204 |
+
|
| 205 |
+
def mel_spectrogram(y, n_fft, num_mels, sampling_rate, hop_size, win_size, fmin, fmax, center=False):
|
| 206 |
+
'''
|
| 207 |
+
if torch.min(y) < -1.:
|
| 208 |
+
print('min value is ', torch.min(y))
|
| 209 |
+
if torch.max(y) > 1.:
|
| 210 |
+
print('max value is ', torch.max(y))
|
| 211 |
+
'''
|
| 212 |
+
global mel_basis, hann_window
|
| 213 |
+
if fmax not in mel_basis:
|
| 214 |
+
#mel = librosa_mel_fn(sampling_rate, n_fft, num_mels, fmin, fmax)
|
| 215 |
+
# sr, n_fft, n_mels=128, fmin=0.0, fmax
|
| 216 |
+
mel = librosa.filters.mel(sr=sampling_rate, n_fft=n_fft, n_mels=num_mels, fmin=fmin, fmax=fmax)
|
| 217 |
+
mel_basis[str(fmax)+'_'+str(y.device)] = torch.from_numpy(mel).float().to(y.device)
|
| 218 |
+
hann_window[str(y.device)] = torch.hann_window(win_size).to(y.device)
|
| 219 |
+
|
| 220 |
+
y = torch.nn.functional.pad(y.unsqueeze(1), (int((n_fft-hop_size)/2), int((n_fft-hop_size)/2)), mode='reflect')
|
| 221 |
+
y = y.squeeze(1)
|
| 222 |
+
|
| 223 |
+
spec = torch.stft(y, n_fft, hop_length=hop_size, win_length=win_size, window=hann_window[str(y.device)],
|
| 224 |
+
center=center, pad_mode='reflect', normalized=False, onesided=True, return_complex=False)
|
| 225 |
+
|
| 226 |
+
spec = torch.sqrt(spec.pow(2).sum(-1)+(1e-9))
|
| 227 |
+
|
| 228 |
+
spec = torch.matmul(mel_basis[str(fmax)+'_'+str(y.device)], spec)
|
| 229 |
+
spec = spectral_normalize_torch(spec)
|
| 230 |
+
|
| 231 |
+
return spec
|
| 232 |
+
|
| 233 |
+
|
| 234 |
+
def get_dataset_filelist_org(a):
|
| 235 |
+
with open(a.input_training_file, 'r', encoding='utf-8') as fi:
|
| 236 |
+
training_files = [os.path.join(a.input_wavs_dir, x.split('|')[0] + '.wav')
|
| 237 |
+
for x in fi.read().split('\n') if len(x) > 0]
|
| 238 |
+
|
| 239 |
+
with open(a.input_validation_file, 'r', encoding='utf-8') as fi:
|
| 240 |
+
validation_files = [os.path.join(a.input_wavs_dir, x.split('|')[0] + '.wav')
|
| 241 |
+
for x in fi.read().split('\n') if len(x) > 0]
|
| 242 |
+
return training_files, validation_files
|
| 243 |
+
|
| 244 |
+
def get_dataset_filelist(a):
|
| 245 |
+
with open(a.input_training_file, 'r', encoding='utf-8') as fi:
|
| 246 |
+
training_files = [x for x in fi.read().split('\n') if len(x) > 0]
|
| 247 |
+
|
| 248 |
+
with open(a.input_validation_file, 'r', encoding='utf-8') as fi:
|
| 249 |
+
validation_files = [x for x in fi.read().split('\n') if len(x) > 0]
|
| 250 |
+
|
| 251 |
+
return training_files, validation_files
|
| 252 |
+
|
| 253 |
+
class MelDataset(torch.utils.data.Dataset):
|
| 254 |
+
def __init__(self, training_files, segment_size, n_fft, num_mels,
|
| 255 |
+
hop_size, win_size, sampling_rate, fmin, fmax, split=True, shuffle=True, n_cache_reuse=1,
|
| 256 |
+
device=None, fmax_loss=None, fine_tuning=False, base_mels_path=None):
|
| 257 |
+
self.audio_files = training_files
|
| 258 |
+
random.seed(1234)
|
| 259 |
+
if shuffle:
|
| 260 |
+
random.shuffle(self.audio_files)
|
| 261 |
+
self.segment_size = segment_size
|
| 262 |
+
self.sampling_rate = sampling_rate
|
| 263 |
+
self.split = split
|
| 264 |
+
self.n_fft = n_fft
|
| 265 |
+
self.num_mels = num_mels
|
| 266 |
+
self.hop_size = hop_size
|
| 267 |
+
self.win_size = win_size
|
| 268 |
+
self.fmin = fmin
|
| 269 |
+
self.fmax = fmax
|
| 270 |
+
self.fmax_loss = fmax_loss
|
| 271 |
+
self.cached_wav = None
|
| 272 |
+
self.n_cache_reuse = n_cache_reuse
|
| 273 |
+
self._cache_ref_count = 0
|
| 274 |
+
self.device = device
|
| 275 |
+
self.fine_tuning = fine_tuning
|
| 276 |
+
self.base_mels_path = base_mels_path
|
| 277 |
+
self.supported_samples = [16000, 22050, 24000] #[4000, 8000, 16000, 22050, 24000, 32000]
|
| 278 |
+
#self.supported_samples = [4000, 8000] #, 16000, 22050, 24000, 32000]
|
| 279 |
+
|
| 280 |
+
def __getitem__(self, index):
|
| 281 |
+
filename = self.audio_files[index]
|
| 282 |
+
while 1:
|
| 283 |
+
#audio, sampling_rate = load_wav(filename)
|
| 284 |
+
audio, sampling_rate = load_segment(filename, self.sampling_rate, self.segment_size)
|
| 285 |
+
if audio is not None: break
|
| 286 |
+
else:
|
| 287 |
+
filename = self.audio_files[random.randint(0,index)]
|
| 288 |
+
#audio, sampling_rate = load_wav(filename)
|
| 289 |
+
#audio, sampling_rate = load_segment(filename, self.sampling_rate, self.segment_size)
|
| 290 |
+
|
| 291 |
+
#audio = audio / MAX_WAV_VALUE
|
| 292 |
+
if not self.fine_tuning:
|
| 293 |
+
audio = normalize(audio) * 0.95
|
| 294 |
+
|
| 295 |
+
sr_out = random.choice(self.supported_samples)
|
| 296 |
+
audio_down = resample(audio, self.sampling_rate, sr_out)
|
| 297 |
+
|
| 298 |
+
target_len = len(audio) #/ downsample_factor
|
| 299 |
+
audio_up = resample(audio_down, None, None, target_len)
|
| 300 |
+
|
| 301 |
+
audio = torch.FloatTensor(audio)
|
| 302 |
+
audio = audio.unsqueeze(0)
|
| 303 |
+
audio_up = torch.FloatTensor(audio_up)
|
| 304 |
+
audio_up = audio_up.unsqueeze(0)
|
| 305 |
+
|
| 306 |
+
mel = mel_spectrogram(audio_up, self.n_fft, self.num_mels,
|
| 307 |
+
self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax,
|
| 308 |
+
center=False)
|
| 309 |
+
|
| 310 |
+
mel_loss = mel_spectrogram(audio, self.n_fft, self.num_mels,
|
| 311 |
+
self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax_loss,
|
| 312 |
+
center=False)
|
| 313 |
+
|
| 314 |
+
return (mel.squeeze(), audio.squeeze(0), filename, mel_loss.squeeze())
|
| 315 |
+
|
| 316 |
+
def __getitem__org(self, index):
|
| 317 |
+
filename = self.audio_files[index]
|
| 318 |
+
if self._cache_ref_count == 0:
|
| 319 |
+
while 1:
|
| 320 |
+
audio, sampling_rate = load_wav(filename)
|
| 321 |
+
if audio is not None: break
|
| 322 |
+
else:
|
| 323 |
+
filename = self.audio_files[random.randint(0,index)]
|
| 324 |
+
audio, sampling_rate = load_wav(filename)
|
| 325 |
+
|
| 326 |
+
audio = audio / MAX_WAV_VALUE
|
| 327 |
+
if not self.fine_tuning:
|
| 328 |
+
audio = normalize(audio) * 0.95
|
| 329 |
+
#self.cached_wav = audio
|
| 330 |
+
if sampling_rate != self.sampling_rate:
|
| 331 |
+
resample_factor = self.sampling_rate / sampling_rate
|
| 332 |
+
new_samples = int(len(audio) * resample_factor)
|
| 333 |
+
audio = scipy.signal.resample(audio, new_samples)#.astype(np.int16)
|
| 334 |
+
#raise ValueError("{} SR doesn't match target {} SR".format(
|
| 335 |
+
# sampling_rate, self.sampling_rate))
|
| 336 |
+
|
| 337 |
+
downsample_factor = 16000 / self.sampling_rate
|
| 338 |
+
new_samples = int(len(audio) * downsample_factor)
|
| 339 |
+
audio_down = scipy.signal.resample(audio, new_samples)
|
| 340 |
+
|
| 341 |
+
new_samples = len(audio) #/ downsample_factor
|
| 342 |
+
audio_up = scipy.signal.resample(audio_down, new_samples)
|
| 343 |
+
#print(f'audio: {audio.shape}, audio_up: {audio_up.shape}')
|
| 344 |
+
#min_idx = min(len(audio), len(audio_up))
|
| 345 |
+
#audio = audio[:min_idx]
|
| 346 |
+
#audio_up = audio_up[:min_idx]
|
| 347 |
+
|
| 348 |
+
self.cached_wav = audio
|
| 349 |
+
self.cached_wav_up = audio_up
|
| 350 |
+
self._cache_ref_count = self.n_cache_reuse
|
| 351 |
+
else:
|
| 352 |
+
audio = self.cached_wav
|
| 353 |
+
audio_up = self.cached_wav_up
|
| 354 |
+
self._cache_ref_count -= 1
|
| 355 |
+
|
| 356 |
+
audio = torch.FloatTensor(audio)
|
| 357 |
+
audio = audio.unsqueeze(0)
|
| 358 |
+
audio_up = torch.FloatTensor(audio_up)
|
| 359 |
+
audio_up = audio_up.unsqueeze(0)
|
| 360 |
+
|
| 361 |
+
if True:
|
| 362 |
+
if self.split:
|
| 363 |
+
if audio.size(1) >= self.segment_size:
|
| 364 |
+
max_audio_start = audio.size(1) - self.segment_size
|
| 365 |
+
audio_start = random.randint(0, max_audio_start)
|
| 366 |
+
audio = audio[:, audio_start:audio_start+self.segment_size]
|
| 367 |
+
audio_up = audio_up[:, audio_start:audio_start+self.segment_size]
|
| 368 |
+
else:
|
| 369 |
+
audio = torch.nn.functional.pad(audio, (0, self.segment_size - audio.size(1)), 'constant')
|
| 370 |
+
audio_up = torch.nn.functional.pad(audio_up, (0, self.segment_size - audio_up.size(1)), 'constant')
|
| 371 |
+
|
| 372 |
+
mel = mel_spectrogram(audio_up, self.n_fft, self.num_mels,
|
| 373 |
+
self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax,
|
| 374 |
+
center=False)
|
| 375 |
+
|
| 376 |
+
mel_loss = mel_spectrogram(audio, self.n_fft, self.num_mels,
|
| 377 |
+
self.sampling_rate, self.hop_size, self.win_size, self.fmin, self.fmax_loss,
|
| 378 |
+
center=False)
|
| 379 |
+
|
| 380 |
+
return (mel.squeeze(), audio.squeeze(0), filename, mel_loss.squeeze())
|
| 381 |
+
|
| 382 |
+
def __len__(self):
|
| 383 |
+
return len(self.audio_files)
|
dataloader/misc.py
ADDED
|
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
#!/usr/bin/env python -u
|
| 3 |
+
# -*- coding: utf-8 -*-
|
| 4 |
+
|
| 5 |
+
from __future__ import absolute_import
|
| 6 |
+
from __future__ import division
|
| 7 |
+
from __future__ import print_function
|
| 8 |
+
import torch
|
| 9 |
+
import torch.nn as nn
|
| 10 |
+
import numpy as np
|
| 11 |
+
import os
|
| 12 |
+
import sys
|
| 13 |
+
import librosa
|
| 14 |
+
import mimetypes
|
| 15 |
+
|
| 16 |
+
def get_file_extension(file_path):
|
| 17 |
+
"""
|
| 18 |
+
Return an audio file extension
|
| 19 |
+
"""
|
| 20 |
+
|
| 21 |
+
_, ext = os.path.splitext(file_path)
|
| 22 |
+
return ext
|
| 23 |
+
|
| 24 |
+
def is_audio_file(file_path):
|
| 25 |
+
"""
|
| 26 |
+
Check if the given file_path is an audio file
|
| 27 |
+
Return True if it is an audio file, otherwise, return False
|
| 28 |
+
"""
|
| 29 |
+
file_ext = ["wav", "aac", "ac3", "aiff", "flac", "m4a", "mp3", "ogg", "opus", "wma", "webm"]
|
| 30 |
+
|
| 31 |
+
ext = get_file_extension(file_path)
|
| 32 |
+
if ext.replace('.','') in file_ext:
|
| 33 |
+
return True
|
| 34 |
+
|
| 35 |
+
mime_type, _ = mimetypes.guess_type(file_path)
|
| 36 |
+
if mime_type and mime_type.startswith('audio'):
|
| 37 |
+
return True
|
| 38 |
+
return False
|
| 39 |
+
|
| 40 |
+
def read_and_config_file(args, input_path, decode=0):
|
| 41 |
+
"""
|
| 42 |
+
Reads and processes the input file or directory to extract audio file paths or configuration data.
|
| 43 |
+
|
| 44 |
+
Parameters:
|
| 45 |
+
args: The args
|
| 46 |
+
input_path (str): Path to a file or directory containing audio data or file paths.
|
| 47 |
+
decode (bool): If True (decode=1) for decoding, process the input as audio files directly (find .wav or .flac files) or from a .scp file.
|
| 48 |
+
If False (decode=0) for training, assume the input file contains lines with paths to audio files.
|
| 49 |
+
|
| 50 |
+
Returns:
|
| 51 |
+
processed_list (list): A list of processed file paths or a list of dictionaries containing input
|
| 52 |
+
and optional condition audio paths.
|
| 53 |
+
"""
|
| 54 |
+
processed_list = [] # Initialize list to hold processed file paths or configurations
|
| 55 |
+
|
| 56 |
+
#The supported audio types are listed below (tested), but not limited to.
|
| 57 |
+
file_ext = ["wav", "aac", "ac3", "aiff", "flac", "m4a", "mp3", "ogg", "opus", "wma", "webm"]
|
| 58 |
+
|
| 59 |
+
if decode:
|
| 60 |
+
if args.task == 'target_speaker_extraction':
|
| 61 |
+
if args.network_reference.cue== 'lip':
|
| 62 |
+
# If decode is True, find video files in a directory or single file
|
| 63 |
+
if os.path.isdir(input_path):
|
| 64 |
+
# Find all .mp4 , mov .avi files in the input directory
|
| 65 |
+
processed_list = librosa.util.find_files(input_path, ext="mp4")
|
| 66 |
+
processed_list += librosa.util.find_files(input_path, ext="avi")
|
| 67 |
+
processed_list += librosa.util.find_files(input_path, ext="mov")
|
| 68 |
+
processed_list += librosa.util.find_files(input_path, ext="MOV")
|
| 69 |
+
processed_list += librosa.util.find_files(input_path, ext="webm")
|
| 70 |
+
else:
|
| 71 |
+
# If it's a single file and it's a .wav or .flac, add to processed list
|
| 72 |
+
if input_path.lower().endswith(".mp4") or input_path.lower().endswith(".avi") or input_path.lower().endswith(".mov") or input_path.lower().endswith(".webm"):
|
| 73 |
+
processed_list.append(input_path)
|
| 74 |
+
else:
|
| 75 |
+
# Read file paths from the input text file (one path per line)
|
| 76 |
+
with open(input_path) as fid:
|
| 77 |
+
for line in fid:
|
| 78 |
+
path_s = line.strip().split() # Split paths (space-separated)
|
| 79 |
+
processed_list.append(path_s[0]) # Add the first path (input audio path)
|
| 80 |
+
return processed_list
|
| 81 |
+
|
| 82 |
+
# If decode is True, find audio files in a directory or single file
|
| 83 |
+
if os.path.isdir(input_path):
|
| 84 |
+
# Find all .wav files in the input directory
|
| 85 |
+
processed_list = librosa.util.find_files(input_path, ext=file_ext)
|
| 86 |
+
else:
|
| 87 |
+
# If it's a single file and it's a .wav or .flac, add to processed list
|
| 88 |
+
#if input_path.lower().endswith(".wav") or input_path.lower().endswith(".flac"):
|
| 89 |
+
if is_audio_file(input_path):
|
| 90 |
+
processed_list.append(input_path)
|
| 91 |
+
else:
|
| 92 |
+
# Read file paths from the input text file (one path per line)
|
| 93 |
+
with open(input_path) as fid:
|
| 94 |
+
for line in fid:
|
| 95 |
+
path_s = line.strip().split() # Split paths (space-separated)
|
| 96 |
+
processed_list.append(path_s[0]) # Add the first path (input audio path)
|
| 97 |
+
return processed_list
|
| 98 |
+
|
| 99 |
+
# If decode is False, treat the input file as a configuration file
|
| 100 |
+
with open(input_path) as fid:
|
| 101 |
+
for line in fid:
|
| 102 |
+
tmp_paths = line.strip().split() # Split paths (space-separated)
|
| 103 |
+
if len(tmp_paths) == 2:
|
| 104 |
+
# If two paths per line, treat the second as 'condition_audio'
|
| 105 |
+
sample = {'inputs': tmp_paths[0], 'condition_audio': tmp_paths[1]}
|
| 106 |
+
elif len(tmp_paths) == 1:
|
| 107 |
+
# If only one path per line, treat it as 'inputs'
|
| 108 |
+
sample = {'inputs': tmp_paths[0]}
|
| 109 |
+
processed_list.append(sample) # Append processed sample to list
|
| 110 |
+
return processed_list
|
| 111 |
+
|